Originally posted by: MichaelPoss

Hi,

How comes solve() can return "Error 1217: No solution exists" ?
(I'm using cplex 12.1 with the JAVA api).

Best,

Michael.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: MichaelPoss

To add a complement of info: exporting my model as a .sav file and trying to solve it tells me it is infeasible, which I expected since it is the subproblem of a Benders decomposition algorithm (and one of the first iterations).
However, I should not get an error by trying to solve an infeasible problem, or am I wrong ?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

> MichaelPoss wrote:
> However, I should not get an error by trying to solve an infeasible problem, or am I wrong ?

That is correct; but if you solve a problem that turns out to be infeasible, and then invoke any function that attempts to access the solution, the objective value etc., you will trigger an error (which I believe is 1217). I think 1217 is also the error you get if you try to access the solution of a feasible problem before you have actually solved it.

Do you invoke IloCplex.getStatus or IloCplex.getCplexStatus immediately after calling solve(), to detect that the problem is infeasible (and avoid attempting to access a nonexistent solution)?

Paul

Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: MichaelPoss

Hi Paul, thank's for your answer.

I agree with you, but the error is already triggered by the solve() call.
In fact, this algorithm has been working fine for a while, that's the first time I get the error.

( After the solve, I ask about the problem's feasibility with isPrimalFeasible() and if it isn't feasible, I call dualFarkas() ).

Best,

Michael.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: LeandroCC

This could be a known issue:

http://www-01.ibm.com/support/docview.wss?uid=swg1RS00991

Have you tried using v. 12.3 or later? Or try the workaround suggested in the link: "one could turn off all heuristics".
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

In case you cannot switch to 12.3 or switching to 12.3 (or higher) does not solve the problem, could you please a full backtrace (or at least the exception's stacktrace up to and including function IloCplex.solve())?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: MichaelPoss

I'll try the switch and let you know.
The heuristics trick does not work.

Michael.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: MichaelPoss

Hi again,

Switching to cplex 12.4 avoids the previous error.
However, a couple of line below I call dualFarkas() which returns the following error:
CPLEX Error 1265: Dual unbounded solution required
Just to be sure, I called isDualFeasible() which tells me that the dual is not feasible, which is consistant with the error returned previously.

My worry comes from the fact that the primal has no objective function. Therefore, the rhs of all constraints in the dual should be 0.
Hence, the dual problem should always be feasible since (0,...,0) is always a solution.
Am I missing something here ?

Michael.

PS: I attach the lp model of the primal.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: T_O

There is no file attached.

Nevertheless, your problem could be due to unbounded variables. You should try to provide upper and lower bounds for every variable. Then your problem should always be dual feasible.

Best regards,
Thomas
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: MichaelPoss

Here comes the file. I prefer not to impose upper bounds on the variables to ease the extraction of the "dual farkas".
However, the constraints of the primal imply upper bounds on all variables.

Best,

Michael.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: T_O

Maybe you can provide some (not too large) trivial upper bounds (or preprocess them manually yourself). Otherwise it might be difficult to ensure dual feasibility.

Best regards,
Thomas
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: T_O

By the way:

http://orinanobworld.blogspot.de/2011/07/farkas-certificates-in-cplex.html

Best regards,
Thomas
#CPLEXOptimizers
#DecisionOptimization